Brief Details: XLS-R-300M fine-tuned for Hausa speech recognition, achieving 36.3% WER with LM. Built on wav2vec2 architecture with Common Voice 8 dataset.
BRIEF DETAILS: MT5-based Chinese question generation model with ROUGE-1 score of 0.4041. Supports text-to-question transformation with easy PyTorch integration and online demo availability.
BRIEF DETAILS: T5-based question generation model fine-tuned on SQuAD 2.0, specializing in creating natural questions from text contexts with 758 downloads and 40 likes.
Brief Details: Ukrainian speech recognition model based on Wav2Vec2-XLSR-53, achieving 32.29% WER on Common Voice dataset. Supports 16kHz audio input.
Brief Details: SciBERT cased model trained on 1.14M scientific papers (3.1B tokens) with custom vocabulary, optimized for scientific text analysis
Brief-details: Longformer Encoder-Decoder (LED) model fine-tuned on arXiv dataset, optimized for scientific paper summarization with 16,384 token context window
Brief Details: MT5-based Portuguese text summarization model fine-tuned for Wiki news, achieving 9.49 ROUGE-1 score. Apache 2.0 licensed with linear learning schedule.
Brief Details: A fine-tuned Wav2Vec2-XLSR-53 model specialized for Kazakh speech recognition, achieving 19.65% WER on test data with 315M parameters.
Brief Details: FinancialBERT - A specialized BERT model pre-trained on 2.4M+ financial documents including news, reports & earning calls, optimized for financial NLP tasks
BRIEF-DETAILS: BART base model fine-tuned on CNN/Dailymail dataset for text summarization. Combines BERT-style encoder with GPT-style decoder. Apache 2.0 licensed.
Brief-details: Financial RoBERTa MLM model (86.6M params) trained on Financial Phrasebank corpus for financial text analysis and prediction tasks
Brief Details: T5-Base model fine-tuned on SQuAD dataset for question generation. 223M parameters, supports both PyTorch and TensorFlow. Popular with 2.3K+ downloads.
Brief Details: A fine-tuned Wav2Vec2-XLSR-53 model specialized for Egyptian Arabic speech recognition, offering automated transcription at 16kHz sampling rate.
BRIEF DETAILS: T5-small-based grammatical error correction model achieving SOTA F0.5 score of 60.70, trained on CLANG-8 and CoNLL datasets.
Brief-details: Finnish BERT-based sentence embedding model (125M params) optimized for paraphrase detection and semantic similarity tasks, supporting Finnish language texts
Brief-details: A 1.3B parameter GPT-Neo model fine-tuned for Vietnamese news text generation, developed by VietAI. Supports causal language modeling with PyTorch implementation.
Brief-details: CPM-Generate is a 2.6B parameter Chinese language model trained on 100GB of diverse text data, capable of text generation, classification, and conversation tasks with strong few-shot performance.
Brief Details: A Czech BERT-like language model optimized for NLP tasks, featuring strong performance in sentiment analysis, NER, and morphological tagging.
Brief-details: A multilingual biomedical named entity recognition model built on mBERT, achieving 98.3% F1 score on biomedical datasets including CRAFT and BC4CHEMD.
BRIEF DETAILS: Indonesian RoBERTa-based emotion classifier with 125M parameters, achieving 72.05% F1-score. Built for Indonesian text emotion analysis using IndoNLU EmoT dataset.
Brief details: Russian BERT-based sentiment analysis model with 178M parameters for classifying text as positive, neutral, or negative. Built for Russian language processing.
